Lexical effects in the production of emotional speech

Abstract

Emotion in speech production has been shown to correlate well with fundamental frequency (F0), intensity, and duration [B. Zupan etal., J.Commun. Disord., 42, 1–17 (2009)]. Studies in non-emotional speech have shown effects of lexical predictors on speech production (e.g., neighborhood density [B. Munson and N. P. Solomon. J. Speech, Lang. Hear. Res., 47, 1048–1058 (2004)], and lexical frequency [R. H. Baayen etal., in Proceedings of the Annual Meeting of the Chicago Linguistic Society (2007), Vol. 43, pp. 1–29]. In the present study, we investigate the effects of lexical predictors in the production of emotional speech. Two professional male actors recorded 260 isolated words. Three emotional states were analyzed: neutral, anger, and joy. Measures of word frequency, morphological family size, number of synonyms and homophones, and ratings for danger and usefulness [L. Wurm, Psychon. Bull. Rev., 14, 1218–1225 (2007)] were used as statistical predictors in modeling, using linear mixed-effects regression. Following previous literature, both speakers use F0, intensity, and duration to portray emotion. Lexical predictors such as word frequency, morphological family size, and danger ratings were found to significantly predict mean F0, mean intensity, and word duration across emotional types for both speakers.